Only in 2.4.20aa1: 00_config-smp-1

	Fixes places using __SMP__ thinking it would work
	as well as CONFIG_SMP.

Only in 2.4.20rc2aa1: 00_drop-inetpeer-cache-2.gz
Only in 2.4.20aa1: 00_drop-inetpeer-cache-3.gz

	Fix smp race due __SMP__ and increase the wrap around
	time per-cpu. from Andi and meissner.

Only in 2.4.20rc2aa1: 00_extraversion-13
Only in 2.4.20aa1: 00_extraversion-14

	Rediffed.

Only in 2.4.20aa1: 00_getcwd-err-1

	Report an error to userspace is getcwd can't return a correct
	value. From Jirka Kosina.

Only in 2.4.20rc2aa1: 00_highio-nohighmem-fix-1
Only in 2.4.20rc2aa1: 00_prepare-write-fixes-3
Only in 2.4.20rc2aa1: 00_presto-dentry-slab-1
Only in 2.4.20rc2aa1: 00_vmalloc-leak-fix-1

	Merged in mainline.

Only in 2.4.20aa1: 00_nfs-xid-smp-1

	Avoid xid races with different mountpoints (in case the highlevel
	locks don't serialize). From Olaf Kirch.

Only in 2.4.20aa1: 00_nfs_writeback-1

	Make sure not to lose writebacks in case of network errors during
	the pageouts from the vm.

Only in 2.4.20aa1: 00_nfsd-reply-cache-smp-1

	More smp locking in nfsd from Olaf Kirch.

Only in 2.4.20aa1: 00_tcp-syn-rst-fin-1

	Don't accept syns if they've rst or fin set too, from Andi Kleen.

Only in 2.4.20rc2aa1: 00_umount-against-unused-dirty-inodes-race-1
Only in 2.4.20aa1: 00_umount-against-unused-dirty-inodes-race-2

	Fix an UP deadlock in 2.4.20rc2aa1.

Only in 2.4.20aa1: 10_sched-o1-bluetooth-1

	o1 sched compile fix, problem noted by Eyal Lebedinsky.
	Eyal, FYI, DAC960 won't compile yet, if you need DAC960
	simply backout elevator-lowlatency until I/somebody will fix it.

Only in 2.4.20aa1: 9925_kmsgdump-0.4.4-1.gz

	Merged kmsgdump, from Willy Tarreau.

Only in 2.4.20rc2aa1: 9942_ocfs-compile-1
Only in 2.4.20aa1: 9942_ocfs-compile-2

	Delete the .c version file.

Only in 2.4.20rc2aa1: 9981_elevator-lowlatency-1
Only in 2.4.20aa1: 9981_elevator-lowlatency-2

	Increase the size to 2M and left unchanged the elvtune settings.  Also
	reduce the cost of a seek, first we don't know if the newly inserted
	request will eventually become big (like a merge), secondly we want
	seeks to be passed over easily by other seeks because the elevator will
	matter more with lots of seeks. We'll see if these changes helps or
	hurts.